Your gateway meter decides how a commercial solar project enters a virtual power plant. A project with its own grid gateway and surplus exports goes in as a generation unit; a project that keeps the array behind the customer meter goes in as load. Neither route runs without an aggregator, and Guangdong allows one generation project per operator.[1] A 10 MW project is still one block inside that unit.
Both routes clear the same bar in Guangdong: a trading unit needs at least 1 MW of adjustable capacity with 1 hour of continuous response, demand-response aggregation needs 5 MW, and every aggregated resource has to sit inside one spot-market clearing node.[1]
The two routes then earn differently. The generation route sells exported kWh at whatever the node clears. A load unit sells verified response against a capped municipal tariff, and only its self-consumed PV gets counted at all.
A PV array moves in one direction only. At noon a 1 MW array can drop about 800 kW within seconds. By 20:00 it can drop 0 kW, so upward regulation needs a battery or an interruptible load.
Start with the money. Guangzhou pays up to 3.5 yuan per kWh of verified active-power response, against 0.25 yuan per kWh for surplus solar on the export tariff, a 14-fold gap.[2]
Two Ways In
Guangdong sorts aggregations into generation and load by the direction electricity flows, not by roof size. Guangzhou targets 5 GW of connected capacity and 1.5 GW of adjustable capacity by the end of 2027, with new distributed PV and user-side storage at 10 kV and above built to be adjustable.[2]
A generation aggregation takes distributed PV, distributed wind and standalone storage that have their own grid-connection gateway at 380 V or 220 V.[1] The generation project registers as the market entity, and the exported kWh are what get sold, settled on the same basis as a new-energy unit.
A load aggregation takes what sits behind the customer meter: self-consumed distributed generation, user-side storage, charging facilities, building air conditioning and interruptible industrial load. Only the self-consumed share of PV counts, and surplus export is capped at 20% of available generation.[1]
Guangdong's registration rules let one operator hold both identities and cap it at 3 spot trading units under a single node,[1] so a site can sell generation and buy load through the same account.
The modules used on this route are N-type 66-cell modules at 615 W to 635 W and 22.8% to 23.5% efficiency. Electrical parameters and mechanical drawings for the 132-cell 2382 mm by 1134 mm frame sit in modules built for commercial and industrial rooftops, alongside the -0.28% per degree Pmax temperature coefficient.
| Item | Generation | Load |
|---|---|---|
| Registers as | Generation project | Electricity user |
| Aggregates | Distributed PV, distributed wind and standalone storage with their own grid gateway | Self-consumed distributed generation, user-side storage, chargers, air conditioning, interruptible load behind the meter |
| Voltage | 380 V or 220 V | The user's existing level |
| Settlement basis | Exported kWh | Change in gateway kWh |
| PV that counts | Surplus export | Self-consumed share |
Where The Limits Sit
Provincial plans set the entry bar. A spot energy trading unit needs 1 MW of adjustable capacity held across 1 continuous hour, and demand-response aggregation needs 5 MW.[1]
Roof area becomes the first filter. A 1 kW array needs about 14 m² of roof against 5 to 6 m² of module area, a ratio of 2.3 to 2.8, so a 1 MW gateway covers roughly 14,000 m². What the self-consumption ratio that picks the model then decides is which role the site qualifies for.
A 200 kW factory roof reaches 20% of that floor, and four more roofs of the same size are needed to form one trading unit.
One constraint bites harder than the rest. Every aggregated resource in a spot trading unit has to sit inside the same clearing node, the same 220 kV bus in Guangdong, and the node rule reorders the shopping list.
Aggregated resources of a virtual power plant taking part in the spot market should in principle sit at the same market clearing node; where grid conditions and market rules allow, cross-node aggregation is also permitted.[3]
| Scenario | Adjustable capacity | Continuous response | Aggregation scope |
|---|---|---|---|
| Spot energy trading unit | 1 MW | 1 hour | One 220 kV clearing node |
| Demand response | 5 MW | 1 hour | Aggregated by city |
How Big The Market Is
China had 470 virtual power plant projects at the end of 2025, carrying 16.85 GW of tested adjustable capacity, up about 70% year on year.[4] The average project holds 35.9 MW, the same order as Guangdong's first 10 MW trading unit.[5]
The 2027 target of 20 GW sits only 18.7% above that 16.85 GW, an annual pace of 8.95% over two years against the 70% that 2025 delivered on its own.[3]
The NDRC and NEA put 20 GW of national adjustable capacity on the board for 2027,[3] and the State Council's carbon peak plan lifts the 2030 figure to 50 GW.[6] Closing the distance from 16.85 GW to 50 GW takes a factor of 2.97, near 24% a year.
By 2030, pumped-storage capacity will reach about 160 GW, new-type storage capacity will aim at 300 GW, and national virtual power plant capacity will exceed 50 GW.[6]
China's cumulative PV fleet reached 1,200 GW in 2025, with 530 GW in distributed systems, 44.2% of the installed total.[7][8] That stock share is a different number from the 42% of PV expansion the IEA puts in distributed applications over 2025 to 2030.[9]
Solar generation hit 1.17 trillion kWh, up 40%, on a 95% utilisation rate.[7] Across 1,200 GW of installed capacity, that 1.17 trillion kWh works out to 975 hours at full power for the year. Midday surplus traces back to that figure.[10]
The US Department of Energy targets 80 GW to 160 GW of virtual power plants by 2030, triple today's scale, for USD 10 billion a year in grid-cost savings.[11] The IEA books 4,600 GW of new renewable capacity between 2025 and 2030, solar near 80%, China near 60%, distributed PV at 42% of the expansion.[10][9]
A PV Array Only Moves One Way
A 1 MW array delivers about 800 kW of downward capacity at noon on a clear day, and 0 kW at 20:00. Divide 1,200 full-load hours by 365 and the daily full-output equivalent is 3.3 hours.
That one-way profile rewrites the revenue shape. Curtailing PV gives up surplus worth 0.25 yuan per kWh, while the same kWh discharged from a battery at the evening peak displaces power bought at 1.0 yuan per kWh, a 4-fold direction gap.
Peak and off-peak arbitrage across a 0.75 yuan/kWh spread is worked through in grid-tied versus hybrid economics against the same 1 MW roof and a 2.2 MWh battery.
A battery closes the gap. A 2.2 MWh system at 80% usable depth over a 2-hour window averages about 880 kW, and that figure sets the bid quantity for upward regulation.
| Asset | Downward | Upward | Available hours |
|---|---|---|---|
| PV only | About 80% of nameplate | 0 kW | Daylight only |
| PV plus a 2.2 MWh battery | About 80% of nameplate | About 880 kW for 2 h | All day |
| PV plus interruptible load | About 80% of nameplate | Per interruptible capacity | Production hours |
The Registration Chain
Registration follows a fixed chain, and a commercial roof walks five steps from signing to its first bid. Block the fifth step, the capability test, and the material from the first four has to be filed again.
- A retail electricity licence comes first, with hardware, software and system integration material filed to the provincial load management centre and the provincial power exchange, covering joint commissioning and security testing.
- The filing goes to public notice and objection handling.
- A resource agency contract follows with the generation project or the electricity user, covering the project list, response pre-notification time, revenue split and assessment sharing.
- Resources are then bound into a trading unit on the exchange system, naming the unit, its type and the products it will trade.
- The trading unit passes a capability test and gains trading status.
A generation trading unit bids the way a new-energy unit does, submitting a 96-point power curve. Spread 96 points across 24 hours and each value covers a quarter hour.
Gateway metering runs on the same beat. Meters freeze every 15 minutes for 35,040 data points a year, the interval behind sizing storage power against storage energy and its 1.8 sizing factor.
Beyond the thresholds, an operator needs a technical support system covering monitoring, forecasting and instruction decomposition, and it has to match the gateway meter's freeze beat.[3] One missing point affects the response assessment for that event.
What Each Service Pays
Guangzhou splits its municipal support into three payments, all settled against verified response volume, out of a city budget capped at 10 million yuan a year.[2]
| Product | Ceiling | Priced on | Nature |
|---|---|---|---|
| Active-power demand response | 3.5 yuan/kWh | Verified response energy | Municipal support |
| Reactive-power demand response | 0.12 yuan/kvarh | Verified response energy | Municipal support |
| Reserve capacity | 0.01 yuan per kW-hour | Carried capacity and hours | Municipal support |
| Spot energy | Cleared at the node | Exported kWh | Market price, not a subsidy |
Take the reserve-capacity payment on its own. A 1 MW unit credited for all 8,760 hours of the year caps out at 87,600 yuan. Credit 3,000 hours instead and it brings 30,000 yuan.
Demand response depends on how often it fires. Use the 880 kW from the table above for 2 hours and one peak event cuts 1,760 kWh, worth 6,160 yuan at the 3.5 yuan/kWh ceiling. Ten events a year come to 61,600 yuan.
Selling the same 17,600 kWh as surplus at 0.25 yuan brings 4,400 yuan, so the gap between the two routes is 14-fold. Demand response plus reserve capacity adds about 91,600 yuan a year before any split, or 91.6 yuan per kW.
Measured against a 2.2 MWh battery, that is a gross return near 4.2%. The two payments stack only when events land inside the same peak window, about 600 hours a year, and the split ratio in the agency contract decides what reaches the site owner.
None of it is free. Reserve capacity is credited by the kW-hour, so 1 MW held for 1 hour is 1,000 kW·h and 10 yuan, and any shortfall against the assessment gets shared out.[3]
The US Department of Energy lists metering and telemetry data-sharing protocols as a precondition for aggregators entering the market, with no double payment for the same service and the aggregator as the single point of contact.[12]
Spot energy is priced at the node, and the same node can register both a generation unit and a load unit. That income comes from the market rather than a capped municipal budget, so it scales with the adjustable capacity a site can carry.
Storage and zero-carbon parks already sit on one line in PV-plus-storage and zero-carbon industrial parks: 56 fishery-PV stations, more than 20 provinces and municipalities, and over 5.1 GW connected.
A Roof Already Doing It
Guangdong's first generation-type virtual power plants entered the market with 5 trading units totalling about 33 MW, an average of 6.6 MW each,[5] all bidding quantity and price into the spot market.
One of them sits in Beijiao, Shunde, where a 47,000 m² industrial park roof was assembled into a single 10 MW trading unit.[5] That works out at 4.7 m² per kW, 0.3 m² per kW below the 5.0 m² module-area benchmark, a 6% tighter fit.
The scale comes from the roof rather than from new plant. Over the same period every distributed source below 110 kV in Guangdong moved into the spot market as a generation-type virtual power plant, taking direct market entry past 33 GW, about 40% of the province's wind and solar fleet.[5]
A 1 MW commercial roof is not unusual, and a commercial rooftop already running in Switzerland sits at the same scale, with about 14,000 m² of roof area to compare against.
Parks and counties hold more. Beyond factory roofs, an agriculture-PV-storage project already operating carries the same kind of adjustable capacity, and one 300 MW site alone could fill 300 units of 1 MW.
Rural roof area runs past 27 billion m², enough in theory for 1,900 GW of rooftop PV, or 3.6 times the 530 GW of distributed solar already installed.[8]
Five Things To Check Before You Sign
Fail any one of these five and the bid never leaves the desk. The industrial and commercial rooftops already built are a useful benchmark set to hold against your own numbers.
- Where the gateway sits: surplus exports go in as generation, full self-consumption goes in as load, and a wrong identity changes every settlement basis.
- Whether adjustable capacity reaches 1 MW: below that, find an operator willing to pool your roof before negotiating terms.
- Whether every resource lands on one clearing node: roofs across a 220 kV bus cannot share a trading unit.
- Whether upward regulation exists at all: without a battery or interruptible load, only curtailment is for sale.
- What the agency contract says about splits and assessment: revenue share, pre-notification time and shortfall sharing belong in writing before signature.
Technical parameters and acceptance items sit in the product specification sheet, and two checks come before any bid: whether the site can produce a power curve, and whether the gateway meter freezes every 15 minutes.
Clear both and the same 1 MW roof collects about 91,600 yuan a year from the virtual power plant side, 91.6 yuan per kW against the site's own consumption and export baseline.
Site assessment
A 1 MW roof plus a 2.2 MWh battery clears the 1 MW / 1 h trading-unit bar on a single gateway meter, so 91,600 yuan a year before the split belongs to one roof instead of five pooled ones. Send us the gateway meter location, the roof area, and the last 12 months of demand peaks, and our engineers will run the same arithmetic against your site before you sign anything.